DAF-MIT AI Accelerator recognizes New York Guardsmen as a key stakeholder in bringing AI capabilities to the fight.

  • Published
  • By Master Sgt. Amber Williams
  • Department of the Air Force AI Accelerator

Cambridge, Mass. — The Department of the Air Force-Massachusetts Institute of Technology Artificial Intelligence Accelerator depends on tactical level Airmen to bring artificial intelligence capabilities to the fight.

On September 21st, Col. Garry Floyd, DAF-MIT AI Accelerator Director, made a surprise appearance at the 174th Attack Wing in Syracuse, New York and presented Capt. Timothy Cullipher the inaugural “AIA Top Stakeholder Award” in person. This highly selective, prestigious award is reserved for Airmen and Guardians who go above and beyond to bring emerging AI technology to the field.

For several years, Cullipher regularly engaged with the Air Force Research Lab in Rome, NY to learn about cutting-edge developments and brainstorm ways to connect research to operations. During this time, Cullipher met Tech. Sgt. Armando Cabrera, a former geospatial intelligence analyst and founding member of the DAF-MIT AI Accelerator. Cabrera and his team at MIT invented an AI model that could generate synthetic color images from radar imagery, enhancing overall readability.

“We don’t have enough people who can properly exploit Synthetic-Aperture Radar imagery,” said Cullipher. “You have to be a pretty well-trained analyst to understand what is going on.” Cabrera’s AI model allowed more operators to analyze the SAR images faster in a way humans are used to reading images, in color.

For AI systems to work well, they require large quantities of usable data and Cullipher saw an opportunity to help the DAF-MIT AI Accelerator advance their research. With support from 174th Attack Wing leadership, Cullipher worked persistently to ensure the data was releasable. “We spent the better part of a year working through intel oversight, making sure we had the proper use memorandums, making sure that we are navigating all the bureaucratic red tape, working with the security manager for unmanned aerial systems, getting the data approved for release to academia to further this research,” said Cullipher. Their relentless efforts allowed the DAF-MIT AI Accelerator to tap into hundreds of gigabytes of data per day.

Capt. Victor “SALSA” Lopez, an MQ-9 pilot and project manager at the DAF-MIT AI Accelerator, took lead for testing Cabrera’s AI model. After ensuring the model worked using MQ-9 SAR data, the race was on to test the model at an upcoming exercise, Valiant Shield 2022.

Utilizing another connection unlocked by Cullipher, the DAF-MIT AI Accelerator partnered with AFRL to use their Agile Condor, an edge computing device which enabled Lopez to run the AI model. It took Lopez two days to load Cabrera’s code and test it on Agile Condor, and only a few minutes to run the code at Valiant Shield once the first SAR images were downloaded.

“The fact that we went from an academic paper to operations in less than four months was mind-blowing and watching how the operators began to use the insights from the model to make relevant tactical decisions made the whole process worth it,” remarked Lopez after the test.

“We’re augmenting human-machine-teaming,” said Cullipher. “You have this scrappy little guard unit out of New York, and just the willingness of everyone to answer emails and all of a sudden we were able to make something happen.”